Back-propagation Feed-forward Artificial Neural Network Research Articles

Spatial quantification method of grassland utilization intensity on the Qinghai-Tibetan Plateau: A case study on the Selinco basin

Existing methods for spatial quantification of grassland utilization intensity cannot meet the demand for accurate detection of the spatial distribution of grassland utilization intensity in the Qinghai-Tibetan Plateau with high spatial resolution. In this paper, a method based on remote-sensing observations and simulations of grassland growth dynamics is proposed. The grassland enhanced vegetation index (EVI) time-series curve during the growing season characterizes the growth of grassland in the corresponding pixel; The deviation between the observed and potential EVI curves indicates the disturbance on grassland growth imposed by human activities, and it can characterize the grassland utilization intensity during the growing season. Based on the main idea described above, absolute and relative disturbances are calculated and used as quantitative indicators of grassland utilization intensity defined from different perspectives. Livestock amount at the pixel scale is obtained by pixel-by-pixel calculations based on the function relationship at the township scale between absolute disturbance and livestock density, which is specific quantitative indicator that considers the mode of grassland utilization. In simulating the potential EVI of grassland, the lag and accumulation effects of meteorological factors are investigated at the daily scale using a multi-objective genetic algorithm. Further, the nonlinear functions between multiple environmental factors (e.g., grassland type, topography, soil, meteorology) and the grassland EVI are established using an error back-propagation feedforward artificial neural network (ANN-BP) with parameter optimization. Finally, the potential EVIs of all grassland pixels are simulated on the basis of this model. The method is applied to the Selinco basin on the Qinghai-Tibetan Plateau and validated by examining the spatial consistency of the results with township-scale livestock density and grazing pressure. The final results indicate that the proposed method can accurately detect the spatial distribution of grassland utilization intensity which is appliable in the similar regions.

The neuromorphological caudate–putaminal clustering of neostriate interneurons: Kohonen self–organizing maps and supervised artificial neural networks with multivariate analysis

AimsThe objective of this study is to investigate the possibility of the neuromorphotopological clustering of neostriate interneurons (NSIN) and their consequent classification into caudate (CIN) and putaminal neuron type (PIN), according to the nuclear localization of the neurons. It tends to discover whether these two topological neuron types are morphologically different. Material and methodsThe binary images of adult human NSIN are used for the purposes of the analysis. The total of the 46 neuromorphological parameters is used. They can be divided into the following classes: neuron surface/size, shape, compartmental length, dendritic branching, neuromorphological organization and complexity. The clustering is performed by an algorithm which consists of the steps of predictor extraction, multivariate cluster analysis set and cluster identification. ResultsUnifactor analysis extracted as significant the following parameters: neurosoma/perikaryon size (AS), the size of a dendritic tree (ADT), the size of a dendritic field area (ADF), the size of an entire neuron field area (ANF), the size of a perineuronal space (APNS), the fractal dimension of a neuron (DN), the index of perikaryon asymmetry (MS), total dendritic length (L), standardized total dendritic length (Lst), standardized dendritic width (DWDTHst), dendritic centrifugal branching order (DCBO), branching polarization index (MDCBO), dendritic partial surface (DSP), the fractal dimension of a skeletonized neuron image (DS), the index of maximal complex density of a dendritic tree (NMAX) and standardized dendritic branching pattern complexity (CDF/ADFst). The cluster analysis set together with Kohonen self–organizing maps and backpropagation feed–forward artificial neural networks confirmed the classification on both unsupervised and supervised manner, respectively. As a final step, the cluster identification is performed by an assignment of each neuron to a particular cluster. ConclusionNSIN can be classified neuromorphologically into CIN and PIN type. Differences are expected since the two nuclei have different functional roles in processing the information involved in volitional movement control.

Minimization of differential column shortening and sequential analysis of RC 3D-frames using ANN

In the preliminary design stage of an RC 3D-frame, repeated sequential analyses to determine optimal members' sizes and the investigation of the parameters required to minimize the differential column shortening are computational effort consuming, especially when considering various types of loads such as dead load, temperature action, time dependent effects, construction and live loads. Because the desired accuracy at this stage does not justify such luxury, two backpropagation feedforward artificial neural networks have been proposed in order to approximate this information. Instead of using a commercial software package, many references providing advanced principles have been considered to code a program and generate these neural networks. The first one predicts the typical amount of time between two phases, needed to achieve the minimum maximorum differential column shortening. The other network aims to prognosticate sequential analysis results from those of the simultaneous analysis. After the training stages, testing procedures have been carried out in order to ensure the generalization ability of these respective systems. Numerical cases are studied in order to find out how good these ANN match with the sequential finite element analysis. Comparison reveals an acceptable fit, enabling these systems to be safely used in the preliminary design stage.

Development and validation of a strain-based Structural Health Monitoring system

An innovative Structural Health Monitoring (SHM) methodology, based on structural strain measurements, which are processed by a back-propagation feed-forward Artificial Neural Network (ANN), is proposed. The demonstration of the SHM methodology and the identification of its capabilities and drawbacks are performed by applying the method in the prediction of fatigue damage states of a typical aircraft cracked lap-joint structure. An ANN of suitable architecture is developed and trained by numerically generated strain data sets, which have been preprocessed by Fast Fourier Transformation (FFT) for the extraction of the Fourier Descriptors (FDs). The Finite Element (FE) substructuring technique is implemented in the stress and strain analysis of the lap-joint structure, due to its efficiency in the calculation of the numerous strain data, which are necessary for the ANN training. The trained network is successfully validated, as it is proven capable to accurately predict crack positions and lengths of a lap-joint structure, which is damaged by fatigue cracks of unknown location and extent. The proposed methodology is applicable to the identification of more complex types of damage or to other critical structural locations, as its basic concept is generic.

Comparative analyses of apple aroma by a tin-oxide gas sensor array device and GC/MS

The apple cultivars “fuji”, “jina” and “huaniu” aroma volatiles were collected and analyzed using a tin-oxide gas sensor array device and the gas chromatography combined with mass spectrometry (GC-MS). Twenty two of the most abundant volatile compounds were taken into account for further study. Eight compounds were found in every cultivar. The principal components analysis (PCA), partial least squares (PLS) and back-propagation feed-forward artificial neural network (BP-ANN) were used to analyze the sensor array and SPME-GC-MS measurements. From the plots of the first two PCs by PCA, different apple cultivars could be clearly distinguished by SPME-GC-MS measurements, while there was slight overlap by sensor array measurements. BP-ANN was used to distinguish different cultivars based on gas sensor array responses, and the accuracy was 87%. Due to the composition of gas sensors in the array, results of PLS models showed that the correlation between fourteen gas sensor array responses and the two PCs of twenty-two compounds were better than the correlation between those and each volatile compound. Furthermore, an ANN was used to build the relationship between the two predicted PCs by PLS model and the three cultivars. The recognition probability was increased to 97%.

Predicting High-Risk Preterm Birth Using Artificial Neural Networks

A reengineered approach to the early prediction of preterm birth is presented as a complimentary technique to the current procedure of using costly and invasive clinical testing on high-risk maternal populations. Artificial neural networks (ANNs) are employed as a screening tool for preterm birth on a heterogeneous maternal population; risk estimations use obstetrical variables available to physicians before 23 weeks gestation. The objective was to assess if ANNs have a potential use in obstetrical outcome estimations in low-risk maternal populations. The back-propagation feedforward ANN was trained and tested on cases with eight input variables describing the patient's obstetrical history; the output variables were: 1) preterm birth; 2) high-risk preterm birth; and 3) a refined high-risk preterm birth outcome excluding all cases where resuscitation was delivered in the form of free flow oxygen. Artificial training sets were created to increase the distribution of the underrepresented class to 20%. Training on the refined high-risk preterm birth model increased the network's sensitivity to 54.8%, compared to just over 20% for the nonartificially distributed preterm birth model.

Artificial neural networks (ANNs) and modeling of powder flow

Effects of micromeritic properties (bulk, tapped and particle density, particle size and shape) on the flow rate through circular orifices are investigated, for three pharmaceutical excipients (Lactose, Emcompress and Starch) separated in four sieve fractions, and are modeled with the help of artificial neural networks (ANNs). Eight variables were selected as inputs and correlated by applying the Spearman product-moment correlation matrix and the visual component planes of trained Self-Organizing Maps (SOMs). Back-propagation feed-forward ANN with six hidden units in a single hidden layer was selected for modeling experimental data and its predictions were compared with those of the flow equation proposed by Jones and Pilpel (1966). It was found that SOMs are efficient for the identification of co-linearity in the input variables and the ANN is superior to the flow equation since it does not require separate regression for each excipient and its predictive ability is higher. Besides the orifice diameter, most influential and important variable was the difference between tapped and bulk density. From the pruned ANN an approximate non-linear model was extracted, which describes powder flow rate in terms of the four network's input variables of the greatest predictive importance or saliency (difference between tapped and bulk density ( x 2), orifice diameter ( x 3), circle equivalent particle diameter ( x 4) and particle density ( x 8)): W=b 0+{a 1[1+ exp(b 2x 2+b 3x 3+b 4x 4+b 8x 8)] −1+a 2[1+ exp(b′ 3χ 3+b′ 8x 8)] −1}.

Prediction of surface hardness after ferritic nitrocarburising of steels using artificial neural networks

A model for prediction of the surface hardness of structural steels after nitrocarburising has been designed. The model is based on a backpropagation feedforward artificial neural network. The performance of the model was checked, using data from the published literature. Good correspondence between predicted from the artificial neural network (ANN) and experimental data was observed. The effects of the chromium, manganese, and nickel contents in the steels on the hardness were studied. Using the model the surface hardnesses after nitrocarburising for some commonly used steels are also predicted. The ANN model has been designed with the aim of optimising the application of nitrocarburising technology.